Blog: Thirst of Pen Caps

Thirst of Pen Caps
Thirst of Pen Caps
Date: 2019-May-03 04:39:02 EST

Today at work, at one of the regular lecture-meetings, the speaker (a relatively young post-doc) provided a really nice framework for understanding machine learning without a set number of layers, with all the right incentives, to go after some aspects of cross-domain learning. I loved it - some years back when I was considering getting a degree in neuroscience at CMU, I thought that cross-domain learning, if I were to enter neuroscience, would be a problem I might devote the rest of my life to. At it turned out I didn't do that, but I always wondered what would've happened if I had (another possible scientific question that's interesting enough for me to have considered this would be to study mechanisms of cell specialisation. It's strange at this later date to have met people who did the specialisation and have insights to share. And sometimes a little sad. I like the path I've taken in my career for the variety, but the path not taken often invokes curiosity.

The other day there was a public lecture I attended - the speaker tried to address the question on whether theories can get at topics that are not directly empirical, the example being whether other universes exist in a multiverse structure (the talk was on physics in counterfactual universes). It was a good talk, but on this question I think he reached the wrong conclusion - he argued that if an existing theory that is itself strongly empirically supported on empirical topics also implies conclusions on these less-empirical topics, it can give us strong reason to reach conclusions on these topics. I believe this is only true if we have good reason to think that the link between those empirical topics and the less-empirical ones is present in all theories that can fit the facts, and consider the space of all possible theories that do so. Without at least a good reason to justify the presence of side-commentary and soft dependencies on those other topics, a successful theory risks capturing more conceptual space than is warranted.

Recently I've been learning that for my own sanity I shouldn't read anything written about MongoDB by people who are still employees there. There's so much misunderstanding of industry and the product by executives of the company (who still like talking about it), while people with more of an engineering bent don't understand industry at all and seem to be really into .. motivated reasoning .. which usually manifests as desperate somewhat-troll-y logic. It can be fun watching someone chew through bad posts, but it's probably best not to read the crud to begin with - there's no value-risk in doing so as I can already make the case for only using Mongo in cases where it's strongly justified, and using Postgres in most other cases.

I think we need to develop a societal consensus (and laws protecting) against the views in this article - while the opening example he raises are fine (work uniform being a voluntary matter), in the general case workers should be able to live their non-work hours as they see fit with no interference (or threat of being fired) from their employer, with only a few obvious restrictions that are avoidable. And if that personal life leads people to pressure the employer, the employer should ignore that pressure (and should in fact be legally required to do so). I would accept a few reasonable limits:

  • Wearing work-provided clothes lessens this independence - employees should wear their own clothes to fully enjoy this independence
  • Attending a conference on the employer's dime deeply lessens this independence
  • Speaking about the employer or other employees largely eliminates this independence, even if it's clear one is not offering official views
  • Speaking about work-relevant notions of fairness or behaviour, particularly when it would imply behaviour relevant to the current workplace, deeply lessens this independence and may require clarification at work (e.g. "I do not trust people of a particular race" suggests issues for interactions with coworkers of that race)
Still, outside limits of these kinds, this is a worker's rights issue and protecting personal lives, even for views that may be reprehensible, is the right thing to do.

I find scorecards to be interesting vehicles of values propogation - Foreign Policy For America is a liberal-leaning org that recently entered the game. Looking over that, I find that I largely (but not entirely) agree with their values. Disagreements and things I'm not sure about:

  • (House) While I support the (past) Iran Deal, I would not think less of people who disagreed with my stances on it
  • (House) I don't think the Hartzler Amendment 183 is a foreign policy matter and I question its inclusion in this metric. I would be mildly supportive of such a measure, as I don't think surgery to help people manage their identity is a good use of military funds.
  • (Senate) I am neutral on what they call the AUMF.
  • (Senate) On Cuba .. they have half a sentence missing in the version of the scorecard that's up as I read this. On this topic, while I support cautious opening of travel ties to Cuba, I think protections and planning should be in place to permit rapid closure of such ties should there be incidents of detention, and a recognition that Cuba's government is despotic leading it to be a dangerous place for American citizens (particularly given our free speech traditions) to travel means we should strucure this opening in ways providing executive flexibility
  • (Senate) On Military Service and the Restore Honor to Service Members Act, I am neutral - I did not like the old policy, but people discharged under it were largely doing so knowingly - I don't see adjusting the outcome to be particularly important. Particularly given that the policy has since been fixed. I'd probably leave it be (open to hear other arguments though)
  • (Senate) On S.1979 which they call the Muslim Ban, at least on its face the actual ban was limited to a few particularly iffy countries, with the only argument as to its unjustifiability being the President's talk on the ban rather than its content. I am wary of taking strong stances on such a legally complex matter (although I condemn the President's talk on that matter)